Bug in incremental propagation

[Andreas Haupt <ahaupt@ifh.de>]

Yes, it's me again ;-)

Our site has three kdc's (one master, two slaves). First the master
database is copied via hprop to the slaves. After that the ipropd-master
is started followed by the start of the ipropd-slave on the slaves.

Now the problems start. When changing the master database only the file
/var/heimdal/log becomes bigger (on master and slaves), the database on
the clients remains unchanged.

So I stopped the ipropd-slave processes and ran the replay_log command.
There I saw this:

ver 3... done
ver 4... replay_log: kadm5_log_replay: No such entry in the database
done
ver 5... done

The error comes from the second last change on the master database which
is already done on the slaves because I copied the database to the slaves
after this change was done.

I added the missing dataset to the slave databases again with kadmin -l
and - replay_log ran fine:

ver 3... done
ver 4... done
ver 5... done
ver 6... done
ver 7... done
ver 8... done

After that I restarted the ipropd-slave processes and /var/heimdal/log
shrank to 24 bytes on the slaves (not on the master). It seems to work
now.

Greetings

-- 
Andreas Haupt         E-Mail: ahaupt@ifh.de
 DESY Zeuthen
 Platanenallee 6
 15738 Zeuthen